Troubleshoot

Described below is a list of issues that may occur under specific situations. We have also provided suggestions on how to diagnose and resolve them quickly.

---

Insufficient Resources

If the host cluster is running out of resources (i.e. compute, storage, memory or GPUs), it can result in downstream issues for the Ray as Service tenants.

Recommendations

Ensure that you plan for sufficient capacity on the host cluster. Configuring alerting and notifications can help significantly with early warning signals.

For example, administrators can use Rafay Kubernetes Manager's integrated monitoring, alerting and proactive email based notifications.

---